Risk Refraction: Thoughts on the Victim-Survivor’s Risk Journey through the Criminal Justice Process

The limits of inter-agency understandings of risk in the context of intimate partner violence are well documented. Informed by Hester’s (2011) ‘three planet’ analogy and using empirical data in one police force area in the south of England, this paper offers an exploration of intra-agency operations, focusing on police risk assessment practices. Exploring the policing risk lens and the victim-survivor journey together, findings highlight police operate with at least three risk assessment moments (call hander, front-line and Safeguarding Hub) and point to the tensions that result when failing to centralise victim-survivors’ own assessment of their risk. Using complexity theory, this paper examines the complex interplay of risk that occurs when the victim-survivor risk journey intersects with the policing aspect of the criminal justice process.


Introduction
Violence against women has an increasing presence in criminal justice policy debates across the globe. The voices of victim-survivors have had a particular impact on policy development in Australia, illustrated by the influence of Rosie Batty and the ongoing public debate generated by the voice of Jess Phillips, among others. Commensurate with a wide range of international criminal justice jurisdictions, Australian policy responses have been suffused and driven by a risk lens (see Cunneen 2014;O'Malley 2004). In relation to intimate partner violence (IPV), criminal justice policy is replete with risk-informed assessment tools as a means of framing response priorities and appropriate resource allocations. These tools are many and varied, with few having been subjected to rigorous empirical validation (McCulloch et al. 2016). Even when validated, research suggests only a limited or modest ability to predict further violence (see Medina, Robinson and Myhill 2016;Turner, Medina and Brown 2019). The commitment to using a risk lens when attempting to keep women safe drives the use of such tools. However, a deeper exploration of this focus on risk measurement tools suggests some problematic limitations in their capacity to secure such safety. Short (1984) observed that the criminal justice policy embrace of risk reflects assumptions concerning what is actionable, measurable and doable. While a measurable understanding of risk can serve the interests of policymakers (Carlen 2008), it stretches the concept of risk in ways that invite limited understandings of the nature of social reality (Mythen 2014). For example, risk assessment tools can conflate prevention with prediction (O'Malley 2006), may not account for the situated social or historical context in which they have been produced or applied (Cunneen 2014), may fail to 'see' gender (Barlow and Walklate 2021) and may ultimately embrace an understanding of risk that is incomplete and partial (Mythen 2014). Fundamentally, whether static or dynamic, risk assessment tools are limited in their capacity to capture the process of risk, as constituted in the lived experiences of women's lives as they exist with, negotiate, and survive the violence they are subjected to. In employing the notion of a 'victimsurvivor journey', the purpose of this paper is to expose some of the co-existing tensions experienced by women and the criminal justice approach to risk. These tensions are experienced as women navigate and are steered through their 'risk journey' within this process.
The paper is comprised of four parts. In the first, following the work of Hester (2011), systematic responses made to IPV are considered within the 'three planet' model. In considering the efficacy of this model, we focus on established problems and possibilities of policing IPV through the risk lens. The second part considers these same systematic responses to IPV, but with a focus on victim-survivor experiences. The concept of the victim-survivor risk journey is introduced, and work associated with viewing the criminal justice process as a journey is reviewed. In the third part, the policing risk lens and the victimsurvivor risk journey are considered in parallel, using empirical data gathered from police officer and victim-survivor experiences of risk assessment drawn from a police force in southern England in 2018. In the concluding part, complexity theory is explored as one way of understanding the complex and complicated interplay that occurs when the victim-survivor intersects with the policing aspect of the criminal justice process. Hester (2011) suggests that despite the positive development of much work to reduce IPV, there remain difficulties and frustrations inhibiting safe outcomes for victim-survivors of domestic abuse. She emphasises the ways in which these difficulties disproportionately impact women and children due to the gendered nature of such violence. She identifies some of the disconnections, tensions, and contradictions evident in professional discourse and practice across three different professional practice arenas: domestic violence (all the agencies dealing with victims and perpetrators), child protection (social work) and child contact (family courts). She concludes that these arenas are especially difficult to bring together in a cohesive and coordinated way because each is essentially 'on a different planet'. She asserts that such a fragmented 'three planet model' precludes effective responses to IPV, and indeed may result in outcomes likely to be counter-productive for individuals interacting with them. She argues for systemic change, orientated towards a cohesive, coordinated, and unified approach across these planets, centring gender both as a feature of IPV and as a feature of service delivery to ensure more effective policy intervention.

The Three Planets (of Risk)
This model, and the insights it affords, reveals much concerning the problems and possibilities of interagency working in IPV and its capacity to be accountable for policy delivery. These problems and possibilities are compounded when the focus of this model is turned towards intra-organisational collaborations, concerned with internal organisational priorities and practices and their application through the organisational process.
Among responses to IPV, much effort has been focused on inter-agency practice and, as a sub-element, particular attention has been paid to the organisational and operational effectiveness of the policing 'planet'. In the context of policing responses to IPV, specific attention has been paid to the variable efficacy of the risk assessment tools used by front-line police officers. This work has identified the problems and possibilities inherent in the policing response to this risk moment. The literature has focused on the potential for improvement in policing practice (Myhill and Hohl 2016) and has highlighted the role described by Robinson et al. (2016) as a 'small constellation of risk factors' in informing decision making. While important, this focus on the role of the front-line officer can direct attention towards this risk moment, despite the initial encounter with a front-line officer only constituting one aspect of a victimsurvivor's risk journey. This focus can also marginalise women's own assessment of their level of risk and how this may or may not be factored into the risk grading(s) assigned by police across the entirety of the risk journey. Indeed, women's ability to predict their own level of risk with a significant degree of accuracy is well documented (see Day et al. 2014;Wheeler and Wire 2014). These issues resonate with the concerns found in an evidence base replete with literature discussing the nature of, and the extent to which, professional judgement acts as an important filter in making risk decisions, and how these filtering practices can play an important part in how the 'evidence' from the tools being used might be subsequently interpreted (see inter alia, Kemshall 2010; Mythen 2011, Westmarland andKelly 2016;Werth 2017).
Thus, much is taken for granted within risk assessment processes, or as Pycroft and Bartollas (2018) might say, becomes 'locked in'. This is often compounded by differing understandings of risk across and within the community of agencies charged with deciding what is actionable and achievable in relation to a particular case, as seen in Hester's (2011) three planet model. Moreover, while multi-agency responses have been central to improving assistance to women living with violence (Robinson 2006), recent work has criticised this approach for its focus on high-risk victims (Myhill and Hohl 2016), with Hamilton et al. (2019) pointing to some of the specific consequences of this. When added to the questionable ability of agencies when engaging in meaningful and effective information sharing (Stanley and Humphreys 2014) and using the same language (Walby et al. 2017), the ineffectiveness of inter-agency working is often tellingly revealed in the outcomes documented by Domestic Homicide Reviews (Dawson 2017).
These observations clearly point to a range of problems when incorporating the concept of risk into practice, in incorporating an appreciation of women's assessment of their own risk into that practice, and the ways in which inter-agency understandings of risk may compound these issues. Notably, little work has documented the role of intra-agency understandings of risk in the IPV victim-survivor's risk journey. There is also a parallel paucity of work discussing this journey per se, and it is to a consideration of these issues that we now turn.

The Victim-Survivor's Risk Journey.
Viewing experiences of the criminal justice process as a journey is not new. The term itself is used to depict several different aspects of the criminal justice experience. Different work has drawn attention to the victims' attritional journey (Walby et al. 2011), their experiential journey (Brooks-Hay et al. 2019) and their narrative journey (Brosi and Rolling 2010; See also Taylor-Dunn, Bowen and Gilchrist 2017). When viewed collectively, this work contributes to a wide body of evidence documenting the mismatch between victim expectations and experiences of criminal justice. This is complemented by the types of cases that are most likely to result in a criminal justice outcome being a court judgement or conviction. Little, if any, international work has focused attention on victim-survivor experiences of the varied risk assessment processes involved in the policing response or placed this in the context of the professionals who perform these risk assessments. This we refer to as the 'risk journey'. Hence the need to place the victim-survivor risk journey side by side with the intra-organisational risk process as articulated by those professionals, predominantly police, involved in it. These considerations form the focus of this paper.

Research Overview
This paper draws on the qualitative analysis of a British Academy funded study conducted with the support of Women's Aid (England) and a policing partner in the south of England. The study involved focus groups with police officers and interviews with victim-survivors of IPV. Ethical approval was granted by Lancaster University's Faculty of Social Sciences Ethics Committee prior to data collection.
The focus groups aimed to explore police officers' understandings of the risks and risk assessment processes associated with IPV. Recruitment for the focus groups involved an email sent to all staff, followed by an invitation to a face-to-face briefing session for those who sought further information. Members of staff who wanted to take part contacted the researchers directly via email. A police force contact facilitated the organisation of the focus groups. Five focus groups with 25 participants in total (22 men and three women) were conducted. Four focus groups were conducted with front-line staff and investigating officers, and one with senior members of staff were asked the same questions about various stages of the policing response. In this paper, each officer has been allocated an individual, anonymous code.
Finally, ten semi-structured interviews with victim-survivors of IPV were conducted. These women were seeking the support of Women's Aid, England, at the time of the interview and Women's Aid supported the research team with the recruitment of participants. This was to ensure that the women were provided with appropriate support and, if required, counselling after the interviews had taken place. The interviews focused attention on these women's experiences of police responses, with a particular emphasis on their understandings of risk. The interviews took a conversational approach and lasted between 1-2 hours. Throughout the text, pseudonyms are used to refer to victim-survivors chosen by each participant. The focus group and interview data were coded and analysed using thematic analysis (Braun and Clarke 2006) to identify overarching themes in the data. To enhance inter-rater reliability, two researchers performed this analytic stage where themes were independently identified within the data and then compared and discussed to reach a consensus. In what follows, we discuss our findings from police officers and victimsurvivors in turn.

The Policing Risk Assessment Journey.
The qualitative data from police officers highlighted at least three identifiable stages of risk assessment typically conducted in cases of IPV: the first point of contact, usually involving a call handler, the front-line officer response, and finally, the Safeguarding Hub or Multi-Agency Safeguarding Hub. The reflections of front-line, investigating and senior officers in relation to each of these risk assessment phases will be discussed in turn.

First Point of Contact: Call Handler Risk Assessment
Usually, the first point in which risk is assessed is when a victim contacts the police and is responded to by a call handler. Most police forces in England and Wales assess risk at this stage by asking the informant questions about the nature and circumstances of the report. This process is often supported by pro forma domestic abuse questions and the use of the THRIVE risk assessment tool 1 . This is a tool that is used for all calls, including calls relating to IPV. The call handler then grades the risk response according to the various categories available to them. These include an immediate response (danger to life or imminent threat of violence), priority response (concern for someone's safety), priority investigation (concern for the individual's welfare, but the risk can be safely managed), scheduled investigation (there is a need for investigation, but this is not time-critical) and initial investigation (can be investigated by phone). Each grade is allocated a different scheduled response time based on the assessed level of risk, namely immediate (within 15 minutes), priority response (within 60 minutes), priority investigation (within eight hours), scheduled response (within three days) and initial investigation (in which a scheduled appointment should be made for officers to complete an initial investigation within three days) (NPCC, 2017). While the response category may be guided by THRIVE, the risk grading is also informed by the professional judgement of the person handling the call.
Focus groups conducted with police officers pointed to some issues and inconsistencies with this call handler risk assessment moment. Across the five focus groups (25 participants in total), ten officers highlighted that the control room/call -handler risk assessment grading had implications for how domestic abuse offences were responded to by front-line response officers. For example, one officer stated: There have been times when I go to a domestic and get there hours after they first called, and the victim just doesn't want to engage. That instantly makes you think I may not put as much effort in, because you know they are not going to support. and Sometimes you are just given minimal information by call handlers, or they say it's one thing like an assault and it's something totally different in reality, it's difficult to know what you're even showing up to sometimes' (FG3, P4).
This quote captures some of the implications of risk gradings allocated by call handlers for the subsequent front-line police response. It also highlights some of the potential problems with the call handler risk moment inaccurately capturing the broader context of IPV cases, as well as issues with both the risk assessment tool and call handlers use of this. Further, the response grading allocated by the call handler risk assessment also appeared to influence officer's perceptions of cases and their outcome. Thus: Risk gradings are important because they usually determine what happens next for that victim. So if someone is medium or standard, they may not be contacted for ages. Or if someone isn't graded as a priority by call handlers, it may take ages for someone to get to them. You know, this stuff is important (FG1, P3).
The trouble is we are pretty much governed by the control room. A lot of the time what you get told on the end of the phone in the control room and what you actually get when you turn up (at the scene) are massively different. If they have recorded something as an assault, you've got to judge that as an assault when you get there, so you're governed by them. So even if there is more going on when you get there, you kind of feel governed by what the control room has said. It's so robotic (FG2, P1).
These quotes suggest that even though officers can exercise their professional judgement when assessing risk at all stages of this process, they felt that the allocated risk grading influenced not only their response time but also their general perception of the overall risks present in the case.
Furthermore, allocation of risk gradings and responses do not occur in a vacuum. The political and economic climate can also shape assessments of risk in IPV cases. The project outlined in this study was conducted at a time when the partner police force was facing the consequences of austerity measures impacting policing more generally in England and Wales. Participants in all focus groups outlined the ways in which resourcing issues impacted control room assessment gradings and subsequent front-line responses. For example: The worst is if you're back on response after a few days off and you're asked to go to a domestic that happened two days ago. 'Cause they haven't had enough units because of resource issues or whatever, and you turn up about an argument that happened days ago, and they immediately don't want to talk to you. They say, "well, where were you two days ago at 3 pm when it all happened", and you know, you're kinda stuck for words (FG2, P1).
There just isn't enough of us. So if a case comes through and it isn't high-risk or high priority, it could take hours for someone to get there. We know it's not good enough, but what can we do? (FG2, P3).
These quotes illustrate the tensions officers felt between needing to respond to victims as quickly as possible and ongoing problems with a shortage of officers 'on the ground' to deal with demand. These tensions became even more salient when considering the further stages in the risk assessment process.

Front-Line Officers: On-Scene DASH Risk Assessments
In England and Wales (as with other jurisdictions), officers are additionally required to conduct a risk assessment when they first attend a domestic abuse case, 'on-scene'. All domestic abuse cases require a pro forma risk assessment to be completed during the first response and the police force in this study used the DASH risk assessment tool 2 . The DASH instrument contains questions addressing issues such as victimsurvivor fear, physical injury, the mental health of the victim and perpetrator, pregnancy, repeat victimisation, escalation of abuse, use of weapons and threats to kill. DASH is graded according to standard, medium and high-risk. Officers are required to assess DASH responses and the context of the incident in informing their risk grading decision. Although professional judgement is a key part of this risk assessment process, concerns have been raised about forces operating a prescriptive system, for example, by counting the number of 'ticks' on the DASH form to assess the level of risk (His Majesty's Inspectorate Commission 2014). The initial write-up of the risk assessment is usually reviewed, in the present force, by the Safeguarding Hub (discussed below). While this outlines the administrative process of risk assessment in practice, it is informed by additional concerns.
The compelling presence of physical violence in relation to risk, commented on by Robinson et al. (2016), was manifested in the focus groups with officers expressing the following views: Until there is physical violence, there isn't really evidence of escalation or serious harm to the victim (FG1, P2).
If it was just an argument, I wouldn't make an arrest. I'd be quite happy to just leave them there. If there is nothing physical and it was just an argument, I would be pretty hacked off if I got arrested for that (FG3, P1).
The ongoing association of physical violence with a higher risk has important implications for the understanding of risk when physical violence is not present. These issues also appeared to influence officers' perceptions of repeat victims in relation to the presence or absence of violence; for example, one officer stated: When you've been out to someone four or five times, you can't help but think what has he got to do for you to leave? In a way, the risk is more on us then. You know, we should be able to say to them you need to sign a memorandum of understanding to say, "I agree I am going to terminate this relationship, I'm gonna move out, change my number etc." because then if that person does get back in touch again and things get worse, and worst-case scenario they are killed, we have something to say we tried and our backs are covered. Maybe then if we started throwing that under the noses of victims, you know, we can't help you unless you help yourself", the ball is in their court then (FG4, P3).
Although this could be reflective of this particular officer's attitudes and beliefs towards domestic abuse, it speaks to wider tensions regarding repeat victimisation, the presence of physical violence and the ways in which risk is interpreted in such contexts. This focus on physical violence, and what is actionable and achievable, is particularly important to consider within the context of limited resources and austerity, as discussed previously and captured below: At the moment, it's just job, job, job, so usually, as soon as you go to a domestic, you just need to do what you need to do, get the DASH done and get out on the next job. We don't really have time to think. These are things we just can't fix in the current climate, unfortunately (FG3, P4).
Such resourcing restrictions are also likely to have a role in the kinds of risk decisions taken at the stage of the Safeguarding Hub, the next stage in planning for and managing a victim's safety.

Risk Assessment and the Safeguarding Hub
The 'Safeguarding Hub' reviews all the safeguarding-related risk assessment and communication forms completed by front-line officers following their arrival at the scene of cases of IPV. Other police officers and staff review the DASH risk assessments carried out at the scene, assess the 'accuracy' of the risk gradings assigned to each case, and re-grade the risk levels where necessary. One officer who worked at the Safeguarding Hub explained that this review process involves a more 'holistic' analysis of the risks present in a case, looking beyond the 'incident' in question and considering the context of the relationship. This holistic analysis involves an examination of the risk factors, some of which might have been overlooked by responding officers, including: cross-referencing the DASH risk assessment responses with the known details of the case; reviewing previous intelligence, any warning markers, incident and crime records or convictions that pertain to the relationship, information about the victim and the suspect. This analysis also includes the context of the case and the nature of the abusive behaviours reported.
The finalised risk status assigned by the Safeguarding Hub outlines the degree of support or safeguarding offered to the victim-survivor. For example, a 'standard-risk' victim would typically receive a text message detailing where they can access domestic abuse support (to which they would have to self-refer), whereas a 'high-risk' victim would typically be contacted and supported in a safeguarding capacity by an agency or a sub-contractor. These might, for example, provide the victim-survivor with personal or domestic security alarms and offer them the services of an Independent Domestic Violence Advisor (IDVA). In addition, their case would be referred to a Multi-Agency Risk Assessment Conference (MARAC) 3 , which would, in turn, identify and action further opportunities for safeguarding and risk-reduction.
However, staff in the Safeguarding Hub also follow additional rubrics directing the process of assessing, communicating, and responding to risk in the context of IPV. Thus, specific contexts can act as 'triggers' for escalating the risk categorisation of a case. These include the seriousness of the crime(s) involved (particularly focusing on the level of violence used and the degree of injury sustained), the case history, particularly if it involves at least three prior domestic abuse-related crimes over the previous 12 months or three or more domestic abuse-related non-crime incidents over the past week. Further factors which may be considered include the victim having recently (typically over the space of the last year) been graded as 'high-risk' in a previous IPV case involving the police, the IPV incident having been witnessed by children, the suspect having a serious and extensive history of IPV with other partners and, relatedly, the victim had previously received a disclosure via the Clare's Law legislation 4 (and thus, the implication is that perpetrator has a history involving domestic abuse). Consequently, there are additional criteria that a victim must meet before she might be awarded or retain the final status of a 'high-risk' victim and receive corresponding safeguarding and support.
One officer working in the Safeguarding Hub spoke about contextual factors that could override the risk format outlined above. For example, Safeguarding Hub staff only typically 'count' previous domestic abuse crimes or incidents involving the couple rather than the perpetrator or the victim individually. This means that the victim might not be considered 'high-risk', despite experiencing repeat victimisation, or if the perpetrator has committed other domestic abuse-related offences (that are not significant 'enough') in the context of other abusive relationships over the past year. Additionally, the Safeguarding Hub normally ensures all cases involving victims who have previously been assessed as 'high-risk' in the past year are graded again in any new incoming cases assessed. However, officers informally indicated that in practice, this would probably only happen if another 'significant' incident -typically as involving serious physical violence, breach of bail conditions or perhaps serious threatening behaviour or sexual offending -had occurred. Similarly, officers said that if a high-risk victim had been the subject of a MARAC relatively recently, it is unlikely a further MARAC would be held following a further high-risk incident unless there had been a 'significant change in circumstances.' Again, as the participant quote demonstrates, many of these contextual caveats can be connected to demand and limited resources: If we (Safeguarding Hub) get a risk assessment and we know it's a repeat, but it's consistently the same kind of level (of abuse), it's very rare we will increase the risk assessment. Only if we see a clear escalation they will tend to go in "the pile", which means we won't really be able to look at it again. Not because we don't want to, it's just capacity (Interview 2).
Finally, it is evident that at the Safeguarding Hub, there is considerable variation in the safeguarding and support made available to victims, depending on their geographic location. For example, in the police force area featured in this research, divisions into regional areas meant that each area was subject to separate bidding and contract negotiations for safeguarding services, with different sub-contractor safeguarding agencies offering different services in different areas. This means that dependent on risk grading and geographical area, victims receive different safeguarding inputs. This discrepancy effectively equates to victims experiencing 'justice by geography', or perhaps more aptly 'safety by geography', echoing the broader regional variation and inequalities in police responses to IPV (Westmarland et al. 2017). As one officer reported: The issue is, we're not joined up. It's really easy to blame call handlers or blame response for not doing things right, but we're often all working off different pages without having a real clue what people across the force are doing (FG1, P3). This kind of inconsistency adds a further layer to the contextual tensions and contradictions that run across and through these three risk systems.
From our analysis, it is possible to see that a victim originally graded as high-risk by a call handler might not be considered high-risk by the front-line officer, and, similarly, a victim graded as high-risk on-scene might not be assessed as high-risk by the Safeguarding Hub. Moreover, in the Safeguarding Hub, high-risk victims do not always receive high-risk safeguarding. This depends on a variety of contextual factors. Medium-risk victims can variably receive safeguarding commensurate with low-, medium-and (if upgraded) high-risk victims, depending on their geographic location, the safeguarding contracts negotiated in these areas and the relative availability of capacity to meet demand. Thus, varying approaches lead to discordant tensions being present in the intra-organisational ways in which risk is assessed within the police. The question remains: how do victim-survivors experience this 'messiness'?

Victim-Survivors Policing Risk Journey
First, it is important to note that in drawing attention to the victim-survivor policing risk journey, this journey does not precisely equate with or map on to the stages outlined above. Victim-survivors often come to their relationship with policing reluctantly, from a diverse range of different prior experiences with policing or, indeed, no experience at all. Further, their experiences of the various interconnected risk assessment processes are often differently understood when compared with the recollections of the police professionals previously discussed. Victim-survivors start their journey in a very different place and with a very different set of concerns. This is well documented in the existing literature. While acknowledging IJCJSD 10(3) 2021 these factors, some of our data from victim-survivors refer to different experiences with call handlers and front-line officers.
For example, on calling the police, Sarah said: Some people are scared to call the police; I know I was. I didn't call the first time he did something; you know what I mean? This was the last resort.
Similarly, Lucy reflects on her stark assessment of her own risk, which prompted her to contact the police: I called the police as a last resort. I knew if I didn't call at this point he could kill me. I didn't want to before then because I know I can usually manage myself. Anyway, I called, said that I thought he was going to kill me and do you know how long it took for them to get to me? Six hours! I just thought, what is the point? Then I thought if I would have just left it, he may have calmed down after another few punches, and I could have gone to bed.
Laura also told us: It took me so much courage to call the police, and I did as a last resort. I told them what was happening to me, and they asked me a load of questions and from this they said my risk wasn't that high. But most of those questions they asked didn't actually allow me to say much about my experiences from my own perspective. How can 20 questions or whatever allow them to determine how safe I am?
At the time of or after being risk assessed and assigned a risk status, victim-survivors also revealed the extent to which their own assessments of their risk went unacknowledged in the responses they received. For example, Jessica was given a safe phone by police and told to use it in an emergency: My friend said he was around the corner, so I used this safe phone as I had been advised to do in an emergency, and I was told off by whoever answered that it wasn't an emergency because he wasn't in the property. But I said he was around the corner, and I was told to ring you if I felt concerned. And I was told to not use it in future, just call 999. But I wouldn't have time to ring 999. I would have had to explain everything to the person on the end of the phone, and they could have sent an officer out, yes, but it would have took too long.
In addition, Lucy recalled: When he was arrested for attempted murder, I told the police that I knew where he would be, the pub on the estate with his friends. And they turned up at my house at around 7 am. I didn't get a phone call until about 5 pm to say they had finally just gone to the pub, and guess what, that's where he was. I knew where he would be. But they didn't listen. I just remember sitting in my house, staring out the window, terrified, waiting for him to come for me, and I was thinking, "please come and make me feel safe" (about the police).

Rumi goes further and suggests:
So it is really important for agencies to look at cases individually. Not just looking at high-risk or low risk. That's what I kept getting told I was low risk. Or the implication that she is Asian, it happens in Asia. But if someone comes to you for help, she needs help. She doesn't see herself as low risk if she has contacted the police. That's the last resort.

She further explains:
They may be dealing with cases like mine all the time, but I don't think they should categorise them as high or low risk. Not everyone will be able to express what's going on in their lives, but if someone is trying to reach out to you for help, then help them 100%. I was never told what risk meant. I didn't know I was classed as high-risk, and my case went to MARAC on one occasion. I didn't know what this meant. Other times I was medium-risk, but there was still the same stuff going on day in day out, but I just maybe hadn't been asked the right questions or wasn't giving the right answers for whatever reason on those days.
Further, one of our respondents (Gemma) recognised the resource and demand issues being faced by front-line officers after having been asked to leave the home she shared with her partner went on to say: They should have arrested him, whether it was the end of their shift and they didn't want the paperwork, or if he was too mouthy and they couldn't be bothered. So, it probably was easier for me to go, but it shouldn't have to be like that. A woman should never have to leave; he should be taken away.
Several of our respondents also offered observations concerning the responses from other professionals from whom they had sought help. These responses recognised issues with understanding the risks associated with IPV with these various agencies. For the victim-survivor, these responses were all part of their experiences and were interconnected. Nicola reported: I did tell my GP, but all she said was, "start putting some money aside so you are able to eventually leave and make an escape", and I was thinking are you mad? I have got no money of my own anyway; he is in charge of all of that. He never let me have anything, I never had control over my money.
Further, Lucy revealed she was pregnant when her partner punched her in the stomach. She did not tell the police about this. When in hospital, she did not want to be sent home, so she said she needed to rest. She said: I begged them to let me stay in hospital, but they wouldn't keep me there. And he was always there, and they would be like "do you want to go home" and I would have to say "yeah of course I do", but I used to make out my pain was worse than it was so they would keep me there. I thought I was being such a fraud; I just didn't want to go home as I was terrified the baby would die.
Rumi also reported her experience with her doctor: So, I went to my GP and she was Asian and she tried to tell me that this stuff happens in Asian households all the time. Had someone flagged at that time to the police or anyone for that matter, maybe I would have been out of this. So that was the first flag. Secondly, nine times I was admitted in hospital for being stressed out at home and they kept saying, "there is something bothering you at home", but they never really pressed this. No one really talked to each other or flagged it up to the doctor, for instance.
In summing up her experience, Nicola reflected how police inaction in her case made her feel disbelieved: I think if I was going through another abusive relationship, knowing the support I can receive now, I would contact someone like Women's Aid. I don't think I would go to the police because I feel it can impact you in wanting to say something if you don't feel believed. If someone has waited so long to say something and finally got the courage, but nothing gets done about it, what message does that send?
Several themes run through these victim-survivor experiences, some of which are highly telling. In summary, these experiences of the policing risk journey speak to various issues, including failures of communication (both inter-and intra-agency) and a failure to appreciate women's assessment of their own risks. These experiences add a further layer of 'messiness' in policy responses to IPV and are powerful illustrations of the limitations inherent in the forensic embrace of risk that has so long held sway on this issue internationally. The question becomes, therefore, how is it possible to make sense of these reported experiences and move responses forward?

Making Sense of Messiness: A Space for Complexity Theory?
Complexity theory focuses attention on not only the notion that human beings and systems are 'complex' but also that the relationship between agency and structure is a duality (a process). In the light of this focus, Room (2016: 3) observes that 'no policy is launched on a greenfield site'. Thus, any policy intervention in any organisational setting enters a tangle of institutional practices that have developed over time. In many ways, this tangle of institutional practices in relation to IPV, often driven by different priorities at different points in time, is what led Hester (2011) to her three planet analogy. Room (2016) argues that within such institutional contexts, individuals re-work, contest and manoeuvre what they do within the world in which they work to make sense of it, especially when faced with the uncertainty of new practices. Much of this 'hunkering down' (Room's term) can be observed in the data presented here and in the ongoing challenge and critique faced by police forces in relation to their response to IPV. There are, however, several further implications of complexity theory that can be used to develop a more nuanced understanding of the policy implementation process (which for the purposes of this paper is taken to be policy responses to IPV). Importantly, complexity theory centres on the view that the policy process is not linear. Rather like relationships suffused with abuse, involving dynamic processes over and through time, organisations are, similarly, process dependent. They can and do adapt and evolve, both internally and as part of the external system of which they are a part.
Complexity theory suggests it is not only human beings and systems that are 'complex' but also that the relationships between individuals and structures are complex. It is, therefore, difficult to abstract the whole system from its parts. While, however, constituent elements might link or interact with one another, one part is not reducible to another. In many ways, therefore, the complex intra-organisational deployment of risk, and its contrast with that of the experiences of victim-survivors observed in the data above, illustrates this very well. The non-linearity of organisational life co-existing with 'hunkering down' mirrors the non-linearity of abuse in relationships. Examples of the hunkering down process include reaching back to the security of the presence of physical violence as a key risk indicator in relation to IPV or engaging in robotic, tick-box approaches to risk assessment. Importantly not all abuse within relationships escalates over time or comes to the attention of the criminal justice process. Awareness of these issues points to the question of the fit between policy response and victim-survivor experiences. As Lewis (2014) has commented in relation to probation service responses to IPV, the failure to grasp the non-linearity of such abuse reflects a failure to understand relationships. If then, intra-organisational relationships are centred, it is easy to appreciate how the risk assessment process is as complex as the sense of risk in women's lives. One way to 'tidy up the messiness' articulated in the data here might be to work better with the knowledge already held by people at different stages in the organisational process (in this case, in relation to risk) and to better understand how they act based on that knowledge.
Pycroft and Bartollas (2018) make a compelling argument for understanding how historical practices in the criminal justice system have woven together concepts of utilitarian power in which risk assessment (particularly pertinent to the discussion here) has become locked in as measurable, doable activities (see also Walklate and Hopkins 2019). This implies putting risk and risk assessment tools in their place as parts of a system inseparable from, but not reducible to, the whole. The key consideration here is that, as constituent parts of the risk system, victim-survivor assessments of risk matter too. Therefore, in addition to committing to improved officer training and the refinement of risk tools, recognising the complexity of victim-survivors risk journeys, and viewing risk and risk assessment tools as constituent parts of a whole system, including victim-survivor voices, are central to the project of change.

Conclusion: Refracting Risk
Our qualitative analysis highlights the fragmented and, at times, conflicting understandings of risk through which police officers and staff interpret and respond to IPV. Indeed, police officers themselves also commented on being conscious of the contradictions across their intra-organisational risk system, which some felt impeded effective responses and interventions in the context of domestic abuse (as one of our participants quoted above clearly articulated). The data cited here illustrate at least three different risk assessment systems, all of which have different ways of grading and assessing risk and with subsequent differing safeguarding implications. This becomes even more complex if we look beyond the context of the police to social work, probation and health, all having their own differing risk assessments processes and which echo Hester's (2011) three planet model. Therefore, in parallel with Hester (2011), this paper has identified disjointed conceptualisations of risk precluding both a consistent and effective response to IPV within one organisation: policing. If the inherent structurally neutral embrace of risk observed in the work of Mythen (2014) and Cunneen (2014) is added to the mix, it is easy to see how factors such as gender, context, and minoritised identities (such as indigeneity, ethnicity, race, migrant status and sexual orientation) might well expose victim-survivors to increased risk of further harm because of the concept of risk itself being deployed. Further, our data have emphasised the importance of organisational and resourcing contexts in shaping police responses across each system, to the risk identified and communicated and, ultimately, to the safeguarding and support offered to victim-survivors. Hence, risk and risk assessment are not 'greenfield sites', to paraphrase Room (2016). Risk is refracted both inter-and intra-organisationally and overrides victim-survivor's assessment of their risk. Arguably, the ways in which risk is refracted along these dimensions need to be better understood before developing any further risk assessment innovations.